Author Affiliations
1National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100101, China2Yunnan Observatories, Chinese Academy of Sciences, Kunming 650217, Yunnan, China3Key Laboratory of Solar Activity, Chinese Academy of Sciences, Beijing 100101, China4University of Chinese Academy of Sciences, Beijing 100049, Chinashow less
Fig. 1. Spectral line profile obtained by fitting of data
Fig. 2. Effect of Doppler velocity of solar autorotation on images at different points of spectral line
Fig. 3. Comparison before and after data processing (wavelengths in nm). (a) Original images; (b) grayscale change of original images; (c) images after edge dimming is removed; (d) grayscale change after edge dimming is removed
Fig. 4. Removing information outside solar circle. (a) Original image;(b) image after polar coordinate transformation;(c) solar part
Fig. 5. PCA decomposition contribution rate
Fig. 6. Image comparison before and after 20th-order reconstruction
Fig. 7. Image pre-processing process
Fig. 8. MLP regression network
Fig. 9. General process of calibration of observing wavelength points using BP neural network
Fig. 10. Correlation diagram of training set and test set for method validation experiment. (a) Training set; (b) test set
Fig. 11. Results of group 1 of grouping test experiment. (a) Correlation diagram of training set;(b) correlation diagram of test set;(c) residual of each data prediction result;(d) variation of residual standard deviation of each day with time
Fig. 12. Variation of residual standard deviation of different groups of grouping test experiment. (a) Group 2; (b) group 3; (c) group 4; (d) group 5
Fig. 13. Test results in range of [-0.006 nm, 0.006 nm]
Hardware | Specification |
---|
CPU | Intel Core i9-9900KF @ 3.60 GHz 8-core | RAM | 64 GB | GPU | NVIDIA GeForce RTX 2070(8 GB) | Mainboard | ASUS TUF Z390-PLUS GAMING |
|
Table 1. Experimental hardware environment
Traditional method | | Machine learning method |
---|
Image acquisition time | Spectral line fitting time | Total time | | Image acquisition time | Time for obtaining corresponding wavelength | Total time | 14-19 min | About 60 s | 15-20 min | | 5 s | 2 s | 7 s |
|
Table 2. Time comparison between traditional calibration method and machine learning calibration method